Data versioning

Data versioning

We suggest to version the data that you upload. This is needed because of the reproducibility of the research that is going to be performed on the datasets. We are suggesting to do this based upon (semantic versioning). A better explanation on using semantic versioning in data can be found here: semantic versioning for data products.

Semantic versioning

The implementation of semantic versioning is as follows.

• Major
  • Amending your dataset with new data
• Minor
  • Correcting errors in the data (e.g. categories which are encoded the wrong way)

Table scheme

The table scheme we use is composed of 2 components, the data version and the table name.

• data-major _ data_minor - tablename

Examples for the ‘core’ tables

• 1_0-non_rep
• 1_0-monthly_rep
• 1_0-trimester_rep
• 1_0-monthly_rep

Examples for the ‘outcome’ tables

• 1_0-non_rep
• 1_0-weekly_rep
• 1_0-monthly_rep
• 1_0-yearly_rep

Other approaches

Semantic versioning is just one of the strategies you can use. Date-based versioning is also a good way to deal with this examples can be:

Examples for the ‘core’ tables

Full datetime

• 2020-11-02 21:36
  • 20201102_2136-non_rep

Date only

• 2020-11-02
  • 202011_02-monthly_rep

Date month - if you have less iterations

• 2020-11
  • 2020-11-trimester_rep

Changelogs

To keep track of all the changed within the different versions of the data you uploaded you can keep track of a changelog.

To view an example, please check out the changelogs of the dictionaries. You can use the same format, but for the data. When you archived the project it becomes less relevant.